The Premier of Alberta is in town promoting the Keystone XL pipeline. She fielded questions at the Brookings Institution, and regarding a question referring to last month’s spill in Arkansas said "these are very isolated incidents, and they don't happen as often as people might suggest that they could."
There are plenty of data on pipeline spills which can be used to examine this claim. Looking at 35 years of data in Alberta collected by Alberta’s Energy and Resources Conservation Board from 1975 to 2010, there were 25,942 company-reported releases from pipelines. Of these, 9,011 released crude oil, diesel oil, oily sludge, synthetic crude, or oil crude bitumen.
Deciphering risk from these data is tricky. Spills are self-reported by the companies, and reporting requirements change, as might compliance with reporting. Also, the frequency of oil spills is a function of how much oil is being transported, something that we do not have data on. The average quantity of oil released per year was much lower in the period 1995-2009, which has been driven by the advances in leak detection, emergency response, and inspection tools, but could also be partly driven by changes in reporting (see graph). For example, spills of less than a barrel were not reported before 1992, and so the drop in the average size of spills after 1992 could be driven by small spills now being included in the sample.
Pipeline Spills in Alberta, 1975-2009 (ERCB)
Nonetheless, examining all the data on spills, including spills of less than a barrel is still useful to shed light on the reasons why pipeline releases occur.
The EPA calls any spill larger than 250 barrels a major spill. Taking the above graph but only including major spills also shows that the number of major spills greater than 250 barrels has been decreasing over time.
Pipeline Spills in Alberta Greater than 250 Barrels, 1975-2009 (ERCB)
However, considering that a spill of 250 barrels 1 barrel would not cause the damage seen in Arkansas, it is important to also focus on larger spills to get an idea of risk. The Arkansas spill is estimated to have been 10,000 barrels. In the Albertan data there were 12 oil spills larger than 10,000 barrels. (The data I have end in mid-2010, but in Alberta there was a 28,000 barrel spill on First Nation land in 2011 and a 22,000 barrel spill in 2012.) So, do 14 spills of this size, over 37 years constitute “very isolated incidents”?
Formally examining these trends, beyond what can be done in a blog post, as well as including years post-2010, would be a worthy exercise. Keystone XL would likely be better-engineered and built than most of the pipelines in this dataset, and leak detection has improved, However, it is important to consider pipeline integrity for decades to come—the Arkansas pipeline was 65 years old. Therefore, for optimal planning, we should not dismiss incidents as “very isolated,” especially since there is still so much to be learned from the data.